Back to EveryPatent.com
United States Patent |
6,055,367
|
Liebmann
,   et al.
|
April 25, 2000
|
Semiconductor device compensation system and method
Abstract
A method is provided that automatically generates compensated semiconductor
devices based on existing VLSI CAD database circuit designs. The preferred
method forms a plurality of edge projection shapes which are intersected
with active area shapes to form gate edge shapes. The gate edge shapes and
residual of the edge shapes are the sorted according to their relative
position. These shapes are then selectively biased according to their
relative position, and then are used to compensate the existing gate
conductor shapes. Thus, this method provides a way to generate gate
structures with compensated gate lengths for n-channel and p-channel
devices based on existing gate, diffusion and implant designs. This system
has the advantage of generating designs with detailed attention to the
placement and minimization of jogs that negatively impact the lithography
performance.
Inventors:
|
Liebmann; Lars Wolfgang (Poughquag, NY);
Sayah; Robert T. (Poughkeepsie, NY)
|
Assignee:
|
International Business Machines Corporation (Armonk, NY)
|
Appl. No.:
|
250909 |
Filed:
|
February 16, 1999 |
Current U.S. Class: |
716/19 |
Intern'l Class: |
G06F 017/50 |
Field of Search: |
395/500.2
|
References Cited
U.S. Patent Documents
4783749 | Nov., 1988 | Duzy et al. | 364/491.
|
5159201 | Oct., 1992 | Frei | 250/492.
|
5289040 | Feb., 1994 | Rogers | 257/666.
|
5369596 | Nov., 1994 | Tokumauru | 364/491.
|
5420800 | May., 1995 | Fukui | 364/491.
|
5477466 | Dec., 1995 | Tripathi et al. | 364/489.
|
Primary Examiner: Teska; Kevin J.
Assistant Examiner: Jones; Hugh
Attorney, Agent or Firm: Schmeiser, Olsen & Watts, Schnurmann; Daniel H.
Parent Case Text
This application is a continuation application of U.S. application Ser. No.
08/781,401, now U.S. Pat. No. 5,877,964, entitled "Semiconductor Device
Compensation System and Method", filed on Jan. 10, 1997, which is hereby
incorporated by reference.
Claims
We claim:
1. A program product comprising:
(A) a compensation mechanism for compensating the length of devices in a
VLSI circuit design that minimizes the number of jogs in the resulting
gate conductor design, the compensation mechanism including mechanisms
for:
a) forming a plurality of edge projections from a plurality of gate
conductor shapes;
b) intersecting said plurality of edge projections with a plurality of
active area shapes, said intersection forming a plurality of gate edge
shapes;
c) subtracting said plurality of gate edge shapes from said plurality of
edge projections, said subtraction forming a plurality of residual edge
shapes;
d) sorting said plurality of residual edge shapes into n-edge, p-edge, n-n,
p-p, n-p and edge-edge shapes;
e) biasing said plurality of gate edge shapes and said residual edge
shapes; and
f) wherein said plurality of gate edge shapes and said residual edge shapes
combine with said gate conductor shapes to result in a compensated device
length; and
(B) computer-readable signal bearing media bearing the compensation
mechanism.
2. The program product of claim 1 wherein the signal bearing media
comprises recordable media.
3. The program product of claim 1 wherein the signal bearing media
comprises transmission media.
4. The program product of claim 1 wherein the compensation mechanism
includes mechanisms for:
sorting said plurality of gate edge shapes into p-channel and n-channel
gate edge shapes; and
elongating said plurality of gate edge shapes.
5. The program product of claim 1 wherein the compensation mechanism
includes mechanisms for:
biasing said p-gate edge shapes, said p-edge residual shapes and said p-p
residual shapes a first predetermined amount;
biasing said n-gate edge shapes, said n-edge residual shapes and said n-n
residual shapes a second predetermined amount; and
biasing said n-p residual shapes.
6. The program product of claim 1 wherein the compensation mechanism biases
said n-p residual shapes by biasing said n-p a third predetermined amount.
7. The program product of claim 1 wherein the compensation mechanism biases
said n-p residual shapes by biasing the p-side of said n-p residual shape
said first predetermined amount and biasing the n-side of said n-p
residual shapes said second predetermined amount.
8. The program product of claim 1 wherein the compensation mechanism
further includes mechanisms for subtracting said edge shapes and said
residual shapes from said gate conductor shapes.
9. A program product comprising:
(A) a compensation mechanism for compensating the length of devices in a
VLSI circuit design, the compensation mechanism including mechanisms for:
a) forming a plurality of edge projections from a plurality of gate
conductor shapes;
b) intersecting said plurality of edge projections with a plurality of
active area shapes, said intersection forming a plurality of gate edge
shapes;
c) sorting said plurality of gate edge shapes into p-channel and n-channel
gate edge shapes;
d) elongating said plurality of gate edge shapes;
e) subtracting said elongated plurality of gate edge shapes from said
plurality of edge projections, said subtraction forming a plurality of
residual edge shapes;
f) sorting said plurality of residual edge shapes into n-edge, p-edge, n-n,
p-p, n-p and edge-edge shapes;
g) biasing said p-gate edge shapes, said p-edge residual shapes and said
p-p residual shapes a first predetermined amount;
h) biasing said n-gate edge shapes, said n-edge residual shapes and said
n-n residual shapes a second predetermined amount;
i) biasing said n-p residual shapes; and
j) subtracting said edge shapes and said residual shapes from said gate
conductor shapes; and
(B) computer-readable signal bearing media bearing the compensation
mechanism.
10. The program product of claim 9 wherein the signal bearing media
comprises recordable media.
11. The program product of claim 9 wherein the signal bearing media
comprises transmission media.
12. The program product of claim 9 wherein the compensation mechanism
biases said n-p residual shapes by biasing said n-p a third predetermined
amount.
13. The program product of claim 9 wherein the compensation mechanism
biases the p-side of said n-p residual shape said first predetermined
amount and biases the n-side of said n-p residual shapes said second
predetermined amount.
14. A program product comprising:
(A) a compensation mechanism for compensating the gate length of devices in
a VLSI circuit design that minimizes the number of jogs in the resulting
gate conductor design, the compensation mechanism comprising:
a) a VLSI database, said VLSI database containing the design levels of said
VLSI circuit design, including an active area design level comprising a
plurality of active area shapes and a gate conductor level comprising a
plurality of gate conductor shapes;
b) a mechanism for forming a plurality of edge projections from said
plurality of gate conductor shapes;
c) an intersection mechanism for intersecting said plurality of edge
projections with a plurality of active area shapes, said intersection
mechanism forming a plurality of gate edge shapes;
d) a subtraction mechanism for subtracting said plurality of gate edge
shapes from said plurality of edge projections, said subtraction mechanism
forming a plurality of residual edge shapes;
e) a sorting mechanism for sorting said plurality of residual edge shapes
into n-edge, p-edge, n-n, p-p, n-p and edge-edge shapes;
f) a biasing mechanism for biasing said p-gate edge shapes, said p-edge
residual shapes and said p-p residual shapes a first predetermined amount,
biasing said n-gate edge shapes, said n-edge residual shapes and said n-n
residual shapes a second predetermined amount, and biasing said n-p
residual shapes; and
g) a subtraction mechanism for subtracting said edge shapes and said
residual shapes from said gate conductor shapes, wherein said plurality of
gate edge shapes and residual edge shapes are combined with said gate
conductor shapes to result in compensated device lengths; and
(B) computer-readable signal bearing media bearing the compensation
mechanism.
15. The program product of claim 14 wherein the signal bearing media
comprises recordable media.
16. The program product of claim 14 wherein the signal bearing media
comprises transmission media.
17. The program product of claim 14 wherein said biasing mechanism bias
said n-p residual shapes a third predetermined amount.
18. The program product of claim 14 wherein said biasing mechanism biases
the p-side of said n-p residual shape said first predetermined amount and
biases the n-side of said n-p residual shapes said second predetermined
amount.
Description
BACKGROUND OF THE INVENTION
1. Technical Field
The present invention generally relates to the manufacture of very large
scale integrated (VLSI) circuits and, more specifically to computer aided
design (CAD) tools used in the generation of circuit patterns.
2. Background Art
A very large scale integrated (VLSI) circuit is typically manufactured on a
silicon wafer by a sequence of material additions (i.e., low pressure
chemical vapor depositions, sputtering operations, etc.), material
removals (i.e., wet etches, reactive ion etches etc.), and material
modifications (i.e., oxidations, ion implants, etc.). The location of
these material additions, removals, and modifications are controlled by a
series of masks. Because of the sheer number of devices on a typical
wafer, the design of the masks are very complex.
Thus, the design of semiconductor devices and the masks used to make the
devices is typically highly automated using sophisticated computer aided
design (CAD) databases and procedures. Typical CAD databases store the
VLSI design as a collection of shapes that are fabricated by a series of
photolithography masks, with each mask part of a design level. A typical
VLSI CAD database can include 20 or more design levels.
One of the more critical elements of a VLSI design is the size the device
gates. Typically the size of the device gates will be controlled by a few
of the multitude of design levels. For example, one design process
simplifies and facilitates complex VLSI design by the automatic generation
of the gate structures for both n-channel and p-channel devices from a
generic gate design level. The generic gate design level is commonly
referred to as the "GC" level for gate conductor commonly used in the gate
structure gates and the mask used to created these shapes is referred to
as the GC mask. A second design level, a diffusion level, defines the
shape of a plurality of diffusion regions. The diffusion level is commonly
referred to as the "AA" level, for "active area" and the masks used to
create the diffusion level are referred to as AA masks. Some diffusion
shapes will become n-type regions and others will become p-type. type GC
and AA designs levels, taken together define the actual gate dimensions,
while many other design levels are used to complete the design.
For example, another design level, typically a "positive-block" and/or
"negative block" implant level determines which shapes in the AA level
will be doped n-type and p-type respectively, and will be therefore be
used to form various drain, source and channel regions.
The automatic generation of the device gates is accomplished by dividing
the generic GC design level into three design levels based on their
geometric interaction with other design levels. More specifically, the
VLSI CAD system defines the gates of the field effect transistors (FETs)
by the intersection of the shapes defining the GC level shapes with the
shapes defining the AA level. The gate structures can then be further
sorted as belonging to either n-channel or p-channel devices by examining
their spatial relationship to the shapes in the positive block or negative
block implant levels.
This system greatly facilitates the generation of complex VLSI devices,
which can easily contain millions of devices. In particular, the use of a
generic GC design level that is divided into areas for n-channel and
p-channel devices, reduces the probability of human design error.
Unfortunately, while the system greatly simplifies the design, the same
system makes it difficult to individually compensate devices without
degrading the resulting design. Device compensation is where the relative
size (i.e., the gate width to length ratio) of some devices is changed,
while other devices are left unchanged or changed in a different way.
There are several reasons why device compensation might be required. For
example, the speed of complementary metal oxide semiconductor (CMOS) field
effect transistors (FETs) is critically dependant on the width to length
ratio of a gate on a given device. In many cases, VLSI circuit design
requires individual compensation of n-channel and p-channel performance to
achieve proper timing between devices. This requires the ability to
selectively compensate the width to length ration of n-channel and
p-channel devices. For example, the timing of logic circuitry critically
depends on accurate speed control on n-channel as well as p-channel
devices, and this may require shortening the length of a n-channel devices
by one amount and shortening the length of all p-channel devices by
another amount.
Thus, as CMOS processes mature the ability to selectively adjust the
physical gate length of n-channel and p-channel devices to compensate for
process and device-physics induced speed differences is needed. In the
prior art, these adjustments are typically accomplished by compensating
the width of lines that define the n-channel and p-channel gate
structures. As stated previously, the gate dimensions are substantially
determined by the GC and AA design levels, and thus those levels must be
modified to selectively compensate the n-channel and p-channel gate
structures.
The operation of a prior art method for device compensation will be
described with reference to FIGS. 8-10. FIGS. 8-10 illustrate two design
levels of a sample VLSI design 1000. Of course, in reality, a VLSI chip
would contain millions of devices and would be far more complex than
illustrated. FIGS. 8-10 illustrate only the gate-conductor (GC) design
level and the active area diffusion (AA) design level, where actual VLSI
chips would include many more additional design levels.
In the VLSI design 1000, the GC level comprises four shapes, 1002, 1004,
1006 and 1008. Likewise, the AA level comprises two shapes 1010 and 1012.
In fabrication, the GC design level will be used to form the gates of the
VLSI device while the AA level will be used to form the VLSI device
diffusion regions. Thus, the GC shapes 1002, 1004, 1006 and 1008 each
represent a conductor gate structure and the AA shapes 1010 and 1012
represent diffusion regions.
A further design level not illustrated in FIG. 8, called an implant level,
determines which AA shapes are p-type and which are n-type diffusion
regions. The implant level could be either a positive or negative implant
block, meaning they respectively define regions that are masked off when a
positive or negative implant is done. In the example of FIG. 8, assume
that either a positive implant block covers AA shape 1010 and/or a
negative implant block covers AA shape 1012 during a positive and/or
negative implant respectively. This will result in AA shape 1010 being a
n-type diffusion region and AA shape 1012 being a p-type diffusion region.
In the design 1000, each overlap of the GC level shape with a AA level
shape forms a field effect transistor. Thus, design 1000 forms six
n-channel devices (GC shapes 1002 and 1004 intersecting with AA shape 1010
twice each, and GC shapes 1006 and 1008 intersecting once with AA shape
1010) and two p-channel devices (GC shapes 1006 and 1008 intersecting with
AA shape 1012).
Current auto generation routines use basic Boolean operations, found in
most CAD routines to first define the intersection of GC and an expansion
of the AA shapes. Turning to FIG. 9, the GC shapes are intersected with
the AA shapes with the resulting shapes comprising the gate regions of the
devices. These shapes are then expanded in the gate width direction,
resulting in the plurality of gate shapes 1020. The expansion of the gate
shapes 1020 is important to ensure adequate overlap of the gate structure
past the actual active area. The overlap is necessary to accommodate
overlay errors, process induced pattern infidelities (corner rounding,
line end shortening, etc.), and diffusion effects.
After the gate shapes arc expanded, they are sorted into gates for
n-channel and p-channel devices. The gate shapes are then intersected with
the appropriate positive block/negative block implant shapes, with the
gate shapes subtracted from the original GC design. This results in shapes
stored in three separate design levels for individual size compensation.
In particular, the gate shapes 1020 for n-channel devices are part of a
n-gate level, the gates shapes 1020 for p-channel devices are part of the
p-gate level, and the remaining CG shapes remain as a residual GC level.
With the GC level divided into n-gate, p-gate and residual GC levels, the
individual devices can be compensated. This is done by taking the n-gate
shapes 1020 and p-gate shapes 1020 and biasing (i.e., narrowing) them
appropriately depending upon the amount of device compensation desired. In
particular, the n-gate shapes 1020 can be narrowed to shorten the gate
length of the n-channel devices and/or the p-gate shapes 1020 can be
narrowed to shorten the gate length of the p-channel devices.
With the n-gate, p-gate and residual GC levels compensated, the three
design levels are added back together. Turning now to FIG. 10, this
results in compensated GC shapes. In particular, the width of the GC lines
that correspond to the gate length of n-channel devices have been reduced
by a predetermined amount and the width of the GC lines that correspond to
the gate length of the p-channel devices have been reduced by a
predetermined amount.
Thus, the prior art allows the gate lengths of the n-channel and p-channel
devices to be selectively adjusted in the context of an auto generation
routine. This method significantly streamlines the chip design process and
helps to eliminate design errors. Unfortunatly, two substantial problems
exist. In particular, this method of compensating gate lengths by
adjusting n-gate and p-gate levels and adding those levels back to the
residual GC level results in GC shapes that have a large number of jogs,
such as jogs 1022. The presence of these jogs has several disadvantages.
First, the jogs significantly increase the CAD data volume. Second, the
presence of the jogs makes defect inspection more difficult, both on the
mask and the final wafers, as it is more difficult to differentiate
between these jogs and some types of defects. Finally, the large number of
jogs combined with their close proximity to the active gate area causes
imaging problems. In particular, significant comer rounding can occur at
the resolution limits of the lithography tool and can cause the image
pattern to change width over a long distance. This results in gate
structures that exhibit continuous, wave like line width variation from
the minimum width of the n-channel gates and p-channel gates to the
maximum width of the poly conductor. This behavior can be extremely
detrimental to device performance.
Some of these problems could in theory be corrected by increasing the GC to
AA overlap, i.e, the amount that the active gates extend past the border
of the diffusion region. In reality, however, one is limited in prior art
methods by the fact that false intersections can resulting from
overextending the AA shapes into such that they intersect with the GC
shapes in undesirable ways. The maximum amount that a gate could be
extended past the border of the diffusion region is therefor governed by
the closest allowable approach of any gate conductor and diffusion shape
without the intent to form a transistor. In high density VLSI devices,
this amount is very small and is thus insufficient to allow sufficient
compensation without excessive jogs.
Thus, there currently exists no method or system for auto generation of
compensated gate shapes that does not result in the excessive jogs in the
resulting design. Therefore, what is needed is an improved method for
device compensation that works in the context of a CAD database and
results in device shapes having minimum jogs.
DISCLOSURE OF INVENTION
According to the present invention, a method is provided to automatically
generate compensated semiconductor devices based on existing VLSI CAD
database circuit designs. The preferred method forms a plurality of edge
projection shapes which are intersected with active area shapes to form
gate edge shapes. The gate edge shapes and residual of the edge shapes are
then sorted according to their relative position. These shapes are then
selectively biased according to their relative position, and then are used
to compensate the existing gate conductor shapes. Thus, this method
provides a way to generate gate structures with compensated gate lengths
for n-channel and p-channel devices based on existing gate, diffusion and
implant designs. This system has the advantage of generating designs with
detailed attention to the placement and minimization of jogs that
negatively impact the lithography performance.
The foregoing and other features and advantages of the invention will be
apparent from the following more particular description of a preferred
embodiment of the invention, as illustrated in the accompanying drawings.
BRIEF DESCRIPTION OF DRAWINGS
The preferred exemplary embodiment of the present invention will
hereinafter be described in conjunction with the appended drawings, where
like designations denote like elements, and:
FIG. 1 is a flow diagram of a preferred method 100;
FIG. 2 is a schematic view of an active area design level and a gate
conductor design level with a plurality of edge projections;
FIG. 3 is a schematic view of an active area design level and a gate
conductor design level with a plurality of gate edge shapes;
FIG. 4 is a schematic view of an active area design level and a gate
conductor design level with a plurality of expanded gate edge shapes;
FIG. 5 is a schematic view of an active area design level and a gate
conductor design level with a plurality of expanded gate edge shapes and a
plurality of residual edge shapes;
FIG. 6 is a schematic view of an active area design level and a gate
conductor design level with a plurality of expanded gate edge shapes and a
plurality of residual edge shapes biased for device compensation;
FIG. 7 is a schematic view of an active area design level and a gate
conductor design level with a compensated gate lengths in accordance with
the preferred embodiment;
FIG. 8 is a schematic view of an active area design level and a gate
conductor design level;
FIG. 9 is a schematic view of an active area design level and a gate
conductor design level and a plurality of gate shapes; and
FIG. 10 is a schematic view of an active area design level and a gate
conductor design level with prior art compensated gate lengths.
BEST MODE FOR CARRYING OUT THE INVENTION
The preferred system and method provides a mechanism to automatically
derive designs for compensated semiconductor devices based on existing
VLSI CAD database circuit designs. The system and method provides a way to
generate gate structures with compensated gate lengths for n-channel and
p-channel devices based on existing gate, diffusion and implant designs.
More specifically, a system and method are provided to generate these
designs with detailed attention to the placement and minimization of jogs
that negatively impact the lithography performance.
Referring now to the drawings, and more particularly to FIG. 1, a flow
diagram illustrating a method 100 for compensating semiconductor gate
length in the context of a VLSI CAD system is illustrated. The operation
of method 100 will be discussed with reference to an example illustrated
in FIGS. 2-7.
The method 100 inputs a VLSI CAD database. The CAD database preferably
includes all the design levels that make up the device. A typical CAD
database can include 20 or more design levels. Of these designs levels, a
few control the gate dimensions of the various devices. In particular, the
gate design level, commonly referred to as the "GC" level for the gate
conductor used to define the gate structures for both n-channel and
p-channel devices. Likewise, a diffusion level of the VLSI design system
defines a plurality of n-type and p-type diffusions. The diffusion level
is commonly referred to as the "AA" level, for "active area." These two
design levels intersected together define the gate dimensions, while many
other design levels are used to complete the design.
Turning briefly to FIG. 8, the GC and AA design levels of a sample VLSI
design 1000 are illustrated. These design levels are the same as prior art
designs with no modifications, and thus the preferred method 100 can be
used with existing VLSI CAD databases.
Again, in the VLSI design 1000, the GC level comprises four shapes, 1002,
1004, 1006 and 1008. Likewise, the AA level comprises two shapes 1010 and
1012. In fabrication, the GC design level will be used to form the gates
of the VLSI device while the AA level will be used to form the VLSI device
diffusion regions. Thus, the GC shapes 1002, 1004, 1006 and 1008 each
represent a conductor gate structure and the AA shapes 1010 and 1012
represent diffusion regions.
A further design level not illustrated in FIG. 8, called an implant level,
determines which AA shapes are p-type and which are n-type diffusion
regions. The implant level could be either a positive or negative implant
block, meaning they respectively define regions that are masked off when a
positive or negative implant is done. In the example of FIG. 8, assume
that either a positive implant block covers AA shape 1010 and/or a
negative implant block covers AA shape 1012 during a positive and/or
negative implant respectively. This will result in AA shape 1010 being a
n-type diffusion region and AA shape 1012 being a p-type diffusion region.
In the design 1000, each overlap of the GC level shape with a AA level
shapes forms a field effect transistors. Thus, design 1000 forms six
n-channel devices (GC shapes 1002, 1004, 1006 and 1008 intersecting with
AA shape 1010) and two p-channel devices (GC shapes 1006 and 1008
intersecting with AA shape 1012).
Turning back to FIG. 1, the first step 102 in the method 100 is to create
edge projections outside all GC level shapes. The edge projections are
preferably formed by forming a rectangle on every edge of the GC shapes.
In particular, each edge is defined by two points, and that edge is
extended out, forming a rectangle. This is done for every edge of GC
shapes 1002, 1004, 1006 and 1008. The edge projections 202 are illustrated
in FIG. 2.
The next step 104 is to intersect the edge projections with the underlying
diffusion shapes. This forms a plurality of shapes called "gate edge
shapes" that locate the edges of all the active gates. Turning to FIG. 3,
the plurality of edge shapes 202 are intersected with the AA shapes 1010
and 1012, forming a plurality of gate edge shapes 302. The gate edge
shapes thus follow the transistor gate along its entire width.
The next step 106 is to sort the gate edge shapes into edge shapes that
correspond to p-channel and n-channel devices (referred to herein as
p-gate and n-gate edge shapes respectively). This sorting is done by
comparing the relative location of the gate edge shapes 302 to the various
diffusion regions. In the illustrated example, AA shape 1010 is a n-type
diffusion shape and hence the gate edge shapes 302 that intersect it are
n-gate edge shapes. Conversely, the AA shape 1012 is a p-type diffusion
shape and the gate edge shapes 302 that intersect it are p-gate edge
shapes. This sorting is illustrated in FIG. 3 by the "P" or "N" adjacent
the relative various edge gate shapes.
The next step 108 is to elongate the gate edge shapes by the minimum GC to
AA overlap. This is done to ensure that all process tolerances can be
contained in the worst case design. In particular, the elongation is
important to ensure adequate overlap of the gate structure past the actual
active area in the event of alignment error. The overlap is necessary to
accommodate overlay errors, process induced pattern infidelities (corner
rounding, line end shortening, etc.), and diffusion effects. The elongated
gate edge shapes 302 are illustrated in FIG. 4.
The next step 110 is to subtract the elongated gate edge shapes from the
original edge projection shapes, forming residual edge shapes. In FIG. 5,
the subtraction of the edge projection shapes 202 (of FIG. 2) from the
elongated gate edge shapes forms a plurality of residual edge shapes
501-540, which are illustrated along with the gate edge shapes 302. Thus,
these residual edge shapes 501-540 are formed using a simple difference
operation.
The next step 112 is to sort the residual edge shapes according to what
type of edge gate shapes and/or GC shape edges they extend between. In
particular, the residual edge shapes are sorted into six groups; i.e.,
shapes that extend from an n-gate edge shape to an GC shape edge (n-edge),
shapes that extend from a p-gate edge shape to an GC shape edge (p-edge),
shapes that extend from an n-gate edge shape to another n-gate edge shape
(n-n), shapes that extend from a p-gate edge shape to another p-gate edge
shape (p-p), shapes that extend from an n-gate edge shape to a p-gate edge
shape (n-p) and shapes that do not touch any gate edge shapes (edge-edge).
Referring back to FIG. 5, residual edge shapes 501, 503, 504, 506, 507,
509, 510, 512, 513, 516, 521 and 522 all extend from a n-gate edge shape
to a GC shape edge. It will be noted that shapes 522 and 516 extend into a
inner GC shape corner while the others will the others terminate at an
outside corner. According to the preferred embodiment this does not make a
difference, and all the above shapes arc sorted into one group as
extending from a n-gate edge shape to a GC shape edge. Likewise, residual
edge shapes 517, 518, 519, 524, 525 and 526 each extend from a p-gate edge
shape to an GC shape edge. The residual shapes 502, 505, 508 and 511
extend from an n-gate edge shape to another n-gate edge shape. The
residual shapes 515 and 523 extend from a n-gate edge shape to a p-gate
edge shape. There are no residual shapes that extend from a p-gate edge
shape to another p-gate edge shape in the example of FIG. 5.
Finally, the residual edge shapes 527, 528, 529, 530, 531, 532, 533, 534,
535, 536, 537, 538, 539 and 540 all do not touch any gate edge shapes, and
are thus sorted into the last group. These shapes are not needed for the
preferred method 100 and thus can be deleted from the database at this
time.
With the residual edge shapes sorted, the next step 114 is to bias the edge
gate shapes and the residual gate shapes according to their various
classifications. This biasing is preferably an expansion of the gate edge
and residual shapes in both directions, such that the various gate edge
shapes and residual shapes overlap the GC shapes a predetermined amount.
As will become clear later, the amount of expansion is determined by the
amount of gate length compensation that is desirable for the various
devices.
In the preferred embodiment, the n-gate edge shapes, the n-edge residual
shapes and the n-n residual shapes are all expanded the same predetermined
amount. Likewise, the p-gate edge shapes, the p-edge residual shapes and
the p-p residual shapes are all expanded the same predetermined amount.
The remaining shapes, the n-p residual shapes, can be treated in several
different ways, with each way having advantages.
One method expands the n-side of the n-p residual shape the same amount as
the other n-shapes, while expanding the p-side of the n-p residual shape
in the same amount as the other p-shapes, with one jog in the middle of
the n-p residual shape where the two sides meet. This method has the
advantage of replacing the multitude of jogs with one jog in the middle of
the GC shape.
A second method either narrows the n-p residual shape or expands the n-p
residual shape a relatively small amount. This results in two jogs, one at
the n-p residual shape/n-gate edge shape junction and the other at the n-p
residual shape/p-gate edge shape junction. This method has the advantage
of improving the ability to visual inspect the device where the expansion
amount for both the n-channel and p-channel gates is very small, although
it does create more jogs than the first method.
If however, the amount of gate length compensation is the same for both
n-channel device gates and p-channel devices gates, all jogs can be
eliminated by expanding the n-p residual shapes by the common expansion
factor.
Turning to FIG. 6, the edge gate shapes and residual shapes are expanded
according to the preferred embodiment. The n-gate edge shapes 302, the
n-edge residual shapes 501, 503, 504, 506, 507, 509, 510, 512, 513, 516,
521 and 522, and the n-n residual shapes 502, 505, 508 and 511 are all
expanded in both directions a predetermined distance. This causes these
shapes to overlap the corresponding GC shapes a set amount.
Likewise, the p-gate edge shapes 302, the p-edge residual shapes 517, 518,
519, 524, 525, and 526, are all expanded in both directions a
predetermined distance (there are no p-p edge residual shapes in the
illustrated example, but if there were they would also be expanded the
same amount). Again this causes these shapes to overlap the corresponding
GC shapes a set amount.
The n-p residual shape 523 has been expanded according to the first method,
and the n-p residual shape 513 has been expanded according to the second
method. In particular, the n-side of the n-p residual shape 523 has been
expanded the same amount as the other n-shapes, while the p-side of the
n-p residual shape 523 has been expanded in the same amount as the other
p-shapes. This method creates one jog 610 in the middle of the n-p
residual shape where the two sides meet. Conversely, the n-p residual
shape 513 has been operated on with the second method and narrowed (or
expanded less) than both the n-shapes and p-shapes.
The next step 116 is to adjust the semiconductor device sizes. This is done
by subtracting the expanded gate edge and residual shapes from the GC
shapes, resulting in narrower GC shapes, which will result in shorter gate
lengths for the various devices.
In the case where the n-p residual shape has been narrowed it will not
overlap and the subtraction of step 116 actually results in the GC shape
being widened at those areas.
Turning to FIG. 7, subtracting the expanded edge shapes from the original
GC shapes results in modified GC shapes that will result in shorter gate
lengths while minimizing the number of jogs in the final design. In
particular, the GC shapes that are used to create only gates for only one
type of device, either n or p channel, such as GC shape 1002 and 1004,
which both are only for n-channel device gates, have been narrowed without
creating any jogs that could result in fabrication problems.
The n-channel gate side and p-channel gate side of GC shape 1006 have both
been narrowed with only one jog 710 in the middle. The elimination of the
other jogs minimizes the risk of fabrication error and makes visual defect
inspection much easier. This method of is preferable where the jog is
large enough to rendered by both the defect inspection tool (typically an
optical tool) and a computer rendering (which also typically have a
limited resolution) of the design. This assures that the jog will not be
confused as a defect. Additionally, this method is preferable where the
jog is small enough such that it will not be rendered by the defect
inspection tool or the computer rendering. In this case, the jog will not
be found nor expected, and thus will not be confused as a defect.
The n-channel gate side and p-channel gate side of GC shape 1008 have also
both been narrowed, but the other areas have been widened (or narrowed
less) and this method creates two jogs 712 and 714 in the GC shape. This
has the advantage of expanding the size of both jogs such that they can
both be rendered by the defect inspection tool and the computer rendering.
Thus, this method is preferable where the size of a single jog (created by
the first method) would have been greater than the inspection tool
resolution and less than the computer rendering tool resolution or vic
versa.
The last step 118 is to create the masks for the resulting design. This
typically is done by sending the CAD data to a specialized mask writer.
The masks can then be used to create VLSI devices will appropriately
biased gate lengths for n-channel and p-channel devices.
Thus, the preferred method 100 is able to compensate the gate lengths of
semiconductor devices in a method that results in minimum jogs in the
resulting shapes. The method 100 can be used with existing VLSI CAD
databases and thus can be used on previous designs without requiring
extensive reworking. The preferred embodiment thus provides a system
method that can automatically adjust the width of n-gate and p-gate shapes
while optimizing the topology of the resulting designs.
While the invention has been particularly shown and described with
reference to a preferred exemplary embodiment thereof, it will be
understood by those skilled in the art that various changes in form and
details may be made therein without departing from the spirit and scope of
the invention.
It is important to note that while the present invention has been described
in the context of a fully functional computer system, those skilled in the
art will appreciate that the mechanisms of the present invention are
capable of being distributed as a program product in a variety of forms,
and that the present invention applies equally regardless of the
particular type of signal bearing media to actually carry out the
distribution. Examples of signal bearing media include: recordable type
media such as floppy disks, hard disks and CD ROMS, and transmission type
media such as digital and analog communication links, including wireless
communication links.
Top